This is an article from 2022, where Gary Marcus made the following predictions and was willing to bet 100 thousand dollars on them:
Here are five predictions, arranged from easiest for most humans to those that would require a high degree of expertise:
1. In 2029, AI will not be able to watch a movie and tell you accurately what is going on (what I called the comprehension challenge in The New Yorker, in 2014). Who are the characters? What are their conflicts and motivations? etc.
2. In 2029, AI will not be able to read a novel and reliably answer questions about plot, character, conflicts, motivations, etc. Key will be going beyond the literal text, as Davis and I explain in Rebooting AI.
3. In 2029, AI will not be able to work as a competent cook in an arbitrary kitchen (extending Steve Wozniak’s cup of coffee benchmark).
4. In 2029, AI will not be able to reliably construct bug-free code of more than 10,000 lines from natural language specification or by interactions with a non-expert user. [Gluing together code from existing libraries doesn’t count.]
5. In 2029, AI will not be able to take arbitrary proofs from the mathematical literature written in natural language and convert them into a symbolic form suitable for symbolic verification.
In the hindsight, it is interesting to see just how much the field was underestimated. But back then, I'd probably be a Marcus-level skeptic too. (I didn't give the idea enough thought in 2024, so I cannot say precisely.)
inglor_cz•52m ago
Here are five predictions, arranged from easiest for most humans to those that would require a high degree of expertise:
1. In 2029, AI will not be able to watch a movie and tell you accurately what is going on (what I called the comprehension challenge in The New Yorker, in 2014). Who are the characters? What are their conflicts and motivations? etc.
2. In 2029, AI will not be able to read a novel and reliably answer questions about plot, character, conflicts, motivations, etc. Key will be going beyond the literal text, as Davis and I explain in Rebooting AI.
3. In 2029, AI will not be able to work as a competent cook in an arbitrary kitchen (extending Steve Wozniak’s cup of coffee benchmark).
4. In 2029, AI will not be able to reliably construct bug-free code of more than 10,000 lines from natural language specification or by interactions with a non-expert user. [Gluing together code from existing libraries doesn’t count.]
5. In 2029, AI will not be able to take arbitrary proofs from the mathematical literature written in natural language and convert them into a symbolic form suitable for symbolic verification.
In the hindsight, it is interesting to see just how much the field was underestimated. But back then, I'd probably be a Marcus-level skeptic too. (I didn't give the idea enough thought in 2024, so I cannot say precisely.)